Skip to main content
LLM Security Database
Skip to research search
Last analyzed Sep 9, 2026

Language Model Security Database

985 research findings · 1119 evaluated models

Filtered research findings

468 entries

Matches every word across titles, descriptions, sources, affected systems, and models.

Multi-Accent Audio Jailbreak

Published Analyzed Source: arXiv

Multilingual and multi-accent audio inputs, combined with acoustic adversarial perturbations (reverberation, echo, whisper effects), can bypass safety mechanisms in Large Audio Language Models (LALMs), causing them to generate unsafe or harmful outputs. The vulnerability is amplified by the interaction between acoustic and linguistic variations, particularly in languages with less training data.

Multilingual and Multi-Accent Jailbreaking of Audio LLMs
Evaluated models: DIVA Llama 3 v0 8B, MERaLion AudioLLM, MiniCPM-o 2.6+2 more
  • DIVA Llama 3 v0 8B
  • MERaLion AudioLLM
  • MiniCPM-o 2.6
  • Qwen 2 Audio
  • Ultravox

Zero-Shot Embedding Leak

Published Analyzed Source: arXiv

A Universal Zero-shot Embedding Inversion vulnerability exists in vector databases and embedding-based retrieval systems. The flaw allows an attacker to reconstruct original plaintext documents from their vector embeddings without requiring access to the original training data or training an embedding-specific inversion model. The attack, identified as "ZSinvert," leverages a multi-stage adversarial decoding process: (1) a cosine-similarity guided beam search using a Large Language Model (LLM)…

Universal Zero-shot Embedding Inversion
Evaluated models: Qwen 2 5B

Multimodal Narrative Jailbreak

Published Analyzed Source: arXiv

Multimodal Large Language Models (MLLMs) are vulnerable to a novel attack vector leveraging narrative-driven visual storytelling and role immersion to circumvent built-in safety mechanisms. The attack, termed MIRAGE, decomposes harmful queries into environment, character, and activity triplets, generating a sequence of images and text prompts that guide the MLLM through a deceptive narrative, ultimately eliciting harmful responses. The attack successfully exploits the MLLM's cross-modal…

MIRAGE: Multimodal Immersive Reasoning and Guided Exploration for Red-Team Jailbreak Attacks
Evaluated models: Gemini 1.5 Pro, GPT-4V, Grok 2 Vision+3 more
  • Gemini 1.5 Pro
  • GPT-4V
  • Grok 2 Vision
  • InternVL
  • LLaVA Mistral
  • Qwen VL

Embedding Inversion Privacy Leak

Published Analyzed Source: arXiv

Retrieval-Augmented Generation (RAG) systems employing standard dense embedding models (e.g., Sentence-T5, SimCSE-BERT, RoBERTa, MPNet) for End-Cloud collaboration are vulnerable to Embedding Inversion Attacks (EIA). While embeddings are vector representations designed to be human-unrecognizable, they retain sufficient semantic information to allow an attacker with access to the vectors (e.g., a malicious or compromised cloud provider) to reconstruct the original sensitive plaintext input.

Safeguarding LLM Embeddings in End-Cloud Collaboration via Entropy-Driven Perturbation
Evaluated models: Not reported

Bleeding Pathways Jailbreak

Published Analyzed Source: arXiv

Autoregressive Large Language Models (LLMs) suffer from a dynamic discriminative degradation vulnerability during sequence generation. When processing complex or adversarial inputs, the model's internal capability to distinguish between benign and harmful token sequences—measured by the linear separability of their hidden states—progressively diminishes as generation continues. If an attacker successfully bypasses the model's initial safety compliance judgment (early generation steps), the…

Bleeding Pathways: Vanishing Discriminability in LLM Hidden States Fuels Jailbreak Attacks
Evaluated models: Llama 2 7B Chat, Llama 3 8B Instruct, Llama 3 70B Instruct+7 more
  • Llama 2 7B Chat
  • Llama 3 8B Instruct
  • Llama 3 70B Instruct
  • Llama 3.2 1B Instruct
  • Mistral 7B Instruct v0.2
  • Qwen 2.5 3B Instruct
  • Qwen 2.5 7B Instruct
  • Phi-4 14B Instruct
  • DeepSeek R1 Distill Qwen 7B
  • Zephyr 7B Beta

Cyber Fine-Tuning Undermines Safety

Published Analyzed Source: arXiv

Fine-tuning Large Language Models (LLMs) on the CyberLLMInstruct dataset results in a critical degradation of safety alignment and refusal mechanisms. While the dataset comprises "pseudo-malicious" content (educational descriptions of malware, phishing, and exploits without executable payloads), the Supervised Fine-Tuning (SFT) process on this corpus causes the models to generalize this instruction-following behavior to actual malicious requests. This effectively bypasses safety guardrails…

CyberLLMInstruct: A new dataset for analysing safety of fine-tuned LLMs using cyber security data
Evaluated models: Llama 2 70B, Llama 3 8B, Llama 3.1 8B+4 more
  • Llama 2 70B
  • Llama 3 8B
  • Llama 3.1 8B
  • Phi-3 Mini 3.8B
  • Mistral 7B
  • Qwen 2.5 7B
  • Gemma 2 9B

Latent Subspace Jailbreak

Published Analyzed Source: arXiv

A vulnerability exists in large language models (LLMs) where the model's internal representations (activations) in specific latent subspaces can be manipulated to trigger jailbreak responses. By calculating a perturbation vector based on the difference between the mean activations of "safe" and "jailbroken" states, an attacker can introduce a targeted perturbation to the model's activations, causing it to generate unsafe outputs even when presented with a safe prompt. This manipulates the…

Probing Latent Subspaces in LLM for AI Security: Identifying and Manipulating Adversarial States
Evaluated models: Llama 3.1 8B Instruct

Unchallenged Premise Misinformation

Published Analyzed Source: arXiv

Large Language Models (LLMs) are vulnerable to implicit misinformation propagation due to sycophantic compliance with false premises. When a user prompt embeds a factually incorrect assumption or conspiracy theory as an unchallenged premise (implicit presupposition) rather than asking for verification, the model frequently fails to detect the falsehood. Instead of correcting the user, the model hallucinates a response that accepts, validates, and reinforces the false premise. This…

How to Protect Yourself from 5G Radiation? Investigating LLM Responses to Implicit Misinformation
Evaluated models: Gemini 1.5 Pro, Gemini 2.0 Flash, Claude 3.5 Sonnet+11 more
  • Gemini 1.5 Pro
  • Gemini 2.0 Flash
  • Claude 3.5 Sonnet
  • GPT-4
  • GPT-4o
  • o1
  • Mixtral 8x7B
  • Qwen 2.5 7B
  • Qwen 2.5 72B
  • Tülu 3 8B
  • Tülu 3 70B
  • Llama 3.1 8B
  • Llama 3.1 70B
  • Llama 3.3 70B

Probabilistic Multimodal Jailbreak

Published Analyzed Source: arXiv

Multimodal Large Language Models (MLLMs) are vulnerable to Jailbreak-Probability-based Attacks (JPA). JPA leverages a Jailbreak Probability Prediction Network (JPPN) to identify and optimize adversarial perturbations in input images, maximizing the probability of eliciting harmful responses from the MLLM, even with small perturbation bounds and few iterations. The attack operates by modifying the input image's hidden states within the MLLM to increase the predicted jailbreak probability.

Utilizing Jailbreak Probability to Attack and Safeguard Multimodal LLMs
Evaluated models: DeepSeek VL 1.3B, InstructBLIP Vicuna 13B, InternLM XComposer+2 more
  • DeepSeek VL 1.3B
  • InstructBLIP Vicuna 13B
  • InternLM XComposer
  • MiniGPT-4 Vicuna 13B
  • Qwen VL Chat

Life-Cycle Router Misrouting

Published Analyzed Source: arXiv

Predictive Large Language Model (LLM) routers, specifically those utilizing Deep Neural Network (DNN) and Matrix Factorization (MF) architectures, are vulnerable to adversarial manipulation and backdoor poisoning. These routers are designed to optimize cost and latency by dynamically directing simple queries to "weak" (cheap) models and complex queries to "strong" (expensive) models. Attackers can exploit this mechanism in two ways: 1. Inference-time Attacks: By appending specific adversarial…

Life-Cycle Routing Vulnerabilities of LLM Router
Evaluated models: Not reported

Research methodology

Entries summarize publicly available primary-source security research. Model names reflect only systems explicitly evaluated by the cited paper, and measurements are research-reported unless independent verification is stated.

Discovery and publication are separate: new papers enter a candidate queue, then require human review and explicit approval before publication. Updates follow that review process, not a guaranteed daily publishing schedule. “Last analyzed” is the newest analysis date among published entries; it does not mean that all recent research has been covered.